[MPlayer-dev-eng] PATCH to support theora-exp in mplayer

[Rich Felker]

On Sat, Apr 23, 2005 at 06:11:38PM +0200, Jindrich Makovicka wrote:
> Timothy B. Terriberry wrote:
> >>I'd be happy with disabling slices in swscaler until someone's willing
> >>to fix them correctly. But I thought someone already did commit a
> >>patch related to this, IIRC..
> > 
> > 
> > If you want a patch without the flip, for testing purposes, see:
> > http://people.xiph.org/~tterribe/mplayer.theora-exp.patch-0.8.2-breaks_swscaler
> > 
> > The test file http://people.xiph.org/~tterribe/mobile_itu601_i_422.ogg
> > exhibits the problems that ensue quite well. The file is in 4:2:2
> > format. With -vo xv, it gets converted to YV12 by swscaler. Only the top
> > and bottom stripes of the current frame are displayed. A few rows from
> > frame (n-1) are displayed below the top stripe, a few rows below that of
> > (n-2), and then the rest contains frame (n-3). Needless to say, it takes
> > 4 frames before you even get a complete picture. Around frame 66, it
> > appears to display a very old buffer for one frame, for no apparent
> > reason. With -vo gl2 or -vo png (converting to 24-bit RGB), things
> > appear to work fine.
> 
> This hack should make swscaler support both top->bottom and bottom->top
> slice directions, please test it. Supporting completely random ordering
> of the slices would (IMO) require rewriting swscaler completely.

IMO not. Just make a separate instance for each slice and keep
boundary lines needed stored somewhere... :)
Still, I agree it's probably not worth it. We need to fix the slice
api so that the filter/vo getting slices will know if they'll come in
some sane order or not, so it can disable slice support if they
won't...

Anyway if this patch works, commit it. :) Then we can remove the
idiotic flip stuff from theora and other codecs.

Rich